Christian Wernecke, Helge Parzyjegla, Gero Mühl, P. Danielis, E. Schweissguth, D. Timmermann
{"title":"使用P4拼接基于内容的发布/订阅通知分布树","authors":"Christian Wernecke, Helge Parzyjegla, Gero Mühl, P. Danielis, E. Schweissguth, D. Timmermann","doi":"10.1109/NFV-SDN50289.2020.9289916","DOIUrl":null,"url":null,"abstract":"Publish/subscribe is a flexible communication pattern for loosely coupled distributed applications. The content-based variant matches each published notification against active subscriptions to determine a set of interested subscribers to which the notification is to be delivered. Since the recipient set can be different for each notification, it is challenging to find and install profitable forwarding rules on the network switches. In this paper, we present novel notification forwarding schemes implemented in P4 that use virtual trees (VTs) installed on switches and additional forwarding information encoded in notification packets that is used to connect VTs, to extend VT branches, or to cut off VT subtrees. For deriving beneficial VTs, we consider (i) topological properties of the physical network, (ii) publisher/subscriber relationships, and (iii) notification statistics. We present a generic algorithm for encoding distribution trees and evaluate our forwarding schemes in a data center network. The results show that our schemes perform well and save network bandwidth by reducing the notification header length.","PeriodicalId":283280,"journal":{"name":"2020 IEEE Conference on Network Function Virtualization and Software Defined Networks (NFV-SDN)","volume":"38 11","pages":"0"},"PeriodicalIF":0.0000,"publicationDate":"2020-11-10","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":"2","resultStr":"{\"title\":\"Stitching Notification Distribution Trees for Content-based Publish/Subscribe with P4\",\"authors\":\"Christian Wernecke, Helge Parzyjegla, Gero Mühl, P. Danielis, E. Schweissguth, D. Timmermann\",\"doi\":\"10.1109/NFV-SDN50289.2020.9289916\",\"DOIUrl\":null,\"url\":null,\"abstract\":\"Publish/subscribe is a flexible communication pattern for loosely coupled distributed applications. The content-based variant matches each published notification against active subscriptions to determine a set of interested subscribers to which the notification is to be delivered. Since the recipient set can be different for each notification, it is challenging to find and install profitable forwarding rules on the network switches. In this paper, we present novel notification forwarding schemes implemented in P4 that use virtual trees (VTs) installed on switches and additional forwarding information encoded in notification packets that is used to connect VTs, to extend VT branches, or to cut off VT subtrees. For deriving beneficial VTs, we consider (i) topological properties of the physical network, (ii) publisher/subscriber relationships, and (iii) notification statistics. We present a generic algorithm for encoding distribution trees and evaluate our forwarding schemes in a data center network. The results show that our schemes perform well and save network bandwidth by reducing the notification header length.\",\"PeriodicalId\":283280,\"journal\":{\"name\":\"2020 IEEE Conference on Network Function Virtualization and Software Defined Networks (NFV-SDN)\",\"volume\":\"38 11\",\"pages\":\"0\"},\"PeriodicalIF\":0.0000,\"publicationDate\":\"2020-11-10\",\"publicationTypes\":\"Journal Article\",\"fieldsOfStudy\":null,\"isOpenAccess\":false,\"openAccessPdf\":\"\",\"citationCount\":\"2\",\"resultStr\":null,\"platform\":\"Semanticscholar\",\"paperid\":null,\"PeriodicalName\":\"2020 IEEE Conference on Network Function Virtualization and Software Defined Networks (NFV-SDN)\",\"FirstCategoryId\":\"1085\",\"ListUrlMain\":\"https://doi.org/10.1109/NFV-SDN50289.2020.9289916\",\"RegionNum\":0,\"RegionCategory\":null,\"ArticlePicture\":[],\"TitleCN\":null,\"AbstractTextCN\":null,\"PMCID\":null,\"EPubDate\":\"\",\"PubModel\":\"\",\"JCR\":\"\",\"JCRName\":\"\",\"Score\":null,\"Total\":0}","platform":"Semanticscholar","paperid":null,"PeriodicalName":"2020 IEEE Conference on Network Function Virtualization and Software Defined Networks (NFV-SDN)","FirstCategoryId":"1085","ListUrlMain":"https://doi.org/10.1109/NFV-SDN50289.2020.9289916","RegionNum":0,"RegionCategory":null,"ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":null,"EPubDate":"","PubModel":"","JCR":"","JCRName":"","Score":null,"Total":0}
Stitching Notification Distribution Trees for Content-based Publish/Subscribe with P4
Publish/subscribe is a flexible communication pattern for loosely coupled distributed applications. The content-based variant matches each published notification against active subscriptions to determine a set of interested subscribers to which the notification is to be delivered. Since the recipient set can be different for each notification, it is challenging to find and install profitable forwarding rules on the network switches. In this paper, we present novel notification forwarding schemes implemented in P4 that use virtual trees (VTs) installed on switches and additional forwarding information encoded in notification packets that is used to connect VTs, to extend VT branches, or to cut off VT subtrees. For deriving beneficial VTs, we consider (i) topological properties of the physical network, (ii) publisher/subscriber relationships, and (iii) notification statistics. We present a generic algorithm for encoding distribution trees and evaluate our forwarding schemes in a data center network. The results show that our schemes perform well and save network bandwidth by reducing the notification header length.